Samsung TV privacy claims center on ads and microphones

By Tom Whitfield

One five-second microphone activation occurred on a Tizen smart television during a Flipper Zero test, according to Hugging Face engineer ngxson. ngxson received no further microphone response over the next several hours. The tester said the test television was already rooted and had several settings disabled, so an unmodified set might behave differently.

What prompted the investigation?

Cybersecurity bounty hunter and reverse engineer log4j noticed unusually specific advertisements on a 2025 Samsung television. The researcher then examined the Tizen operating system and live-posted the investigation on X, saying the code had revealed: “discovered PRIVACY HORROR.”

The posts centered on code and observed behavior that could support advertising profiles across connected devices. log4j identified cross-device identification as the central mechanism, describing an AID, a tracking identifier linked to advertising, alongside telemetry gathered from network packets, audio categories and television screenshots.

The account characterized the system as trying to extract as much user information as possible to improve targeted ads. The practical takeaway is simple: the claimed commercial purpose is better targeted advertising, while the report offers no revenue measurement.

What did Tizen reportedly process?

log4j proposed that Samsung televisions respond slowly because they perform substantial post-processing of audio, screenshots and AID data. The account also described computer-vision processing and other models running on the television itself.

The researcher presented that performance explanation as a theory. The post included no timing tests. log4j also found a region-specific convolutional neural network, identified in the post as an AI system.

According to the account, it tracked whether viewers watched particular videos, skipped advertisements and how far they progressed through videos. The model therefore covered both content choices and reactions to ads. The practical limit is clear: the code analysis indicates on-device behavior profiling, while the report offers no advertising revenue or model-performance measurements.

How did connected microphones enter the process?

For televisions without built-in microphones, log4j described code that searches for connected devices with microphones, such as a soundbar or a connected Bluetooth device, to record audio. That path turns an external accessory into a potential microphone for the television. It also changes the privacy boundary because the television can obtain room audio through a device the user bought for sound.

ngxson tested the behavior with a Flipper Zero and observed the television open the microphone for five seconds before stopping, then waited for another microphone response but found none hours later. ngxson said the test television was already rooted and had multiple settings disabled, so an unmodified television may respond differently.

The reported test supports one short activation under those conditions, not persistent microphone use across unmodified Samsung televisions.

Why was the researcher’s OpenAI account banned?

The account described using OpenAI tools to hack the television. OpenAI later banned log4j for cyber abuse, according to an email received by the researcher. log4j denied using those tools to exploit systems or find vulnerabilities, saying they served only for local reverse engineering and did not breach the terms of use.

In a follow-up comment, the researcher said the ban appeared to follow from a primitive classifier. The account of the research includes no response from OpenAI explaining the decision. The practical distinction is clear: the account ban concerns research methods, not evidence about Samsung’s code.

What remains open?

Other connected-television brands have faced similar allegations, including complaints involving LG. That comparison remains allegation-level and does not establish equivalent behavior. The published account included no response from Samsung or an independent auditor.

Television owners have also reported slower performance despite having no new features or heavy applications running. Some suspect accumulated bloatware; log4j’s findings offer another possible explanation, but the report does not establish that privacy processing caused the slowdowns.

log4j is preparing a comprehensive technical write-up that would explain the method and show others how to reproduce the tests. The final post in the series joked that the researcher had “finally found a company that truly listens to its customers.” The coming write-up should show whether the findings reproduce on an unmodified television and how Samsung’s approach compares with the wider connected-TV advertising market.